Predicting asthma phenotypes: characterization of IL1RL1 in asthma
Dijk, Fokelina Nicole
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2018
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Dijk, F. N. (2018). Predicting asthma phenotypes: characterization of IL1RL1 in asthma. Rijksuniversiteit Groningen.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Phenotypic and functional
translation of IL1RL1 locus
polymorphisms in lung
tissue and airway epithelium
in asthma
_
Chapter 4
F. Nicole Dijk*, Michael A. Portelli*, Nick Shrine, Jenny Hankinson, Sangita Bhaker, Marie. E. Ketelaar, Yize Wan, Néomi S. Grotenboer, Ma’en Obeidat, Amanda P. Henry, Charlotte K. Billington, Dominick Shaw, Simon Johnson, Zara E.K. Pogson, Andrew Fogarty, Tricia M. McKeever, David C. Nickle, Yohan Bossé, Maarten van den Berge, Alen Faiz, Sharon Brouwer, Judith M. Vonk, Paul de Vos, Amisha Singapuri, Liam Heaney, Adel H. Mansur, Rekha Chaudhuri, Neil C. Thomson, John W. Holloway, Gabrielle A. Lockett, Peter H. Howarth, Robert Niven, Angela Simpson, John D. Blakey, Martin D. Tobin, Dirkje S. Postma, Ian P. Hall, Louise Wain, Martijn C. Nawijn, Christopher E. Brightling, Gerard H. Koppelman†, and Ian Sayers†
*These authors contributed equally to this work as shared first authors †These authors contributed equally to this work as shared senior authors
Abstract
Background
The IL1RL1 (ST2) gene locus has been reproducibly associated with asthma. However, the contribution of this locus to specific asthma-subtypes remains undefined. Therefore, we tested association of IL1RL1 region SNPs to subtypes of asthma as defined by clinical and immunological measures and addressed their func-tional effects in lung tissue and airway epithelium.
Methods and results
Utilising three independent cohorts and resequencing data, we identified six key IL1RL1 locus signals that drive association with multiple asthma-subtypes, with the most significant signals being associated with blood eosinophil counts, atopy and childhood-onset asthma. Investigations in lung tissue and primary bronchial epithelial cell cultures identified context-dependent relationships between the six key SNPs and expression of different IL1RL1 mRNA isoforms and soluble ST2 protein. Asthma bronchial epithelial cell cultures exposed to exacerbation-relevant stimulations (IL33, Rhinovirus 16 and House Dust Mite) revealed modulatory effects for three SNPs (rs13431828, rs1420101 and rs12465392) towards IL1RL1 mRNA and protein expression suggesting SNP-environment interactions. A four amino acid changing haplo-type in the IL1RL1 TIR domain tagged by rs10192157, affected IL33 driven NF-Kβ signalling, whilst not in-terfering with Toll-like receptor signalling.
Conclusion
In summary, multiple independent mechanisms regulated by different SNPs may explain the complexity of the IL1RL1 region as an association signal in asthma GWAS.
Introduction
Asthma is a chronic airway disorder characterized by chronic inflammation and widespread variable air-flow obstruction that is often reversible, either spontaneously or with treatment.1 Over the years, a
signif-icant genetic component to asthma has been identified and today, over 130 single nucleotide polymor-phisms (SNPs) have been reported to be associated with asthma and allergic diseases in Genome-Wide Association Studies (GWAS).2-4 One of the most replicated asthma associated genetic regions is the
chro-mosome 2q12 locus containing the Interleukin-1 receptor like 1 (IL1RL1), IL18R1 and IL18RAP genes.3,5-7 IL-1RL1 is predominantly expressed as two major splice variants, one of which contains the transmembrane
domain encoding the membrane bound receptor (ST2L, IL1RL1-b). This IL33 receptor is expressed on a number of different cell types relevant to asthma, including inflammatory cells such as T-lymphocytes, innate lymphoid cells, basophils, eosinophils and mast cells, as well as structural cells such as fibroblasts, endothelial and epithelial cells.8,9. An alternate splice variant encodes the soluble form of the receptor
(sST2, IL1RL1-a), which has been detected in both bronchiolar lavage (BAL) and serum. This splice variant is hypothesised to act as a decoy receptor for its ligand, IL33.10,11
The presence of multiple SNPs in the IL1RL1 locus that independently contribute to asthma risk com-plicates the interpretation of the association signal with the disease.4,12 As asthma is known to be a
multi-factorial and heterogeneous disease1, we hypothesise that SNPs within the IL1RL1 locus may have
distinct functional effects, with different SNPs relating to different sub-types or components of asthma. Disease-associated SNPs may exert their functional effects by changing the protein sequence and/or by affecting levels of gene transcription (expression Quantitative Trait Locus, (eQTL)). Whereas some SNPs affect gene expression under constitutive conditions (constitutive eQTL), it has recently been shown that the effect of a SNP on gene transcription is sometimes observed only in a specific context, such as dis-eased conditions (inducible eQTL).13 We hypothesize that the genetic heterogeneity of the IL1RL1 locus
may partly be due to inducible eQTLs that affect gene transcription in asthma, but not in healthy controls. In this study we set out to extend the association of the IL1RL1 region polymorphisms with asthma diagnosis and to define the relative contribution of SNPs spanning the association signal, to subtypes of asthma defined by clinical and immunological measures. To approach this hypothesis we used a step-wise study method to ultimately prioritising selected association signals for functional charac-terisation (Figure 1).
Figure 1. Flow-Diagram of different stages of investigation carried out in this study.
Following detection of known common variation in the locus, we identified coding and non-coding rare variation through resequencing of the IL1RL1 locus in two geographically distinct European populations of asthma patients, to provide a clearer picture of genetic variation in the region. We subsequently re-lated these SNPs to different asthma-subtypes, to identify key priority SNPs for functional investiga-tion. We tested the presence of specific eQTLs in lung and airway bronchial epithelium with a focus to
IL1RL1 regulation, and assessed their role in regulating epithelial IL1RL1 expression after stimulation with
known asthma exacerbation factors, such as human rhinovirus 16 (RV16), a known modulator of IL33 ex-pression14, house dust mite (HDM) and in an artificially IL33 rich environment. Moreover, we performed
reductionist functional studies to address the effect of coding SNPs in IL1RL1 on IL33 induced signal trans-duction. In the same system we investigated the effect of IL1RL1 coding SNPs on Toll-like receptor (TLR)-2 and -4 signalling, which have previously been linked to ST2-TLR crosstalk in the context of tolerance.15,16
Overall, we provide greater insight into the specific sub-phenotypes of asthma that may be driven by genetic variants in this locus including; childhood onset asthma and eosinophilic asthma. Importantly, we provide functional translation of key SNP effects in the lung, including the identification of novel en-vironment-SNP interactions that regulate IL1RL1 expression and activity of relevance to asthma.
Results
Demographics
Three cohorts were investigated: Genetics of Severe Asthma Phenotypes (GASP), Dutch Asthma Genetics (DAG)17 and the Manchester Asthma and Allergy Study (MAAS).18 In the adult GASP cohort, genotypes
for 2,536 cases were available for analyses. This presented with a mean subject age of 47.8 years. Mean age of asthma onset was 23.17 years with 46.2% being diagnosed with asthma before the age of 16 years and 46.8% of the cases were atopic. In the DAG cohort, 909 cases were available with genetic data, with a mean age of 34.7 years. Mean age of asthma onset in this cohort was 10.07 years with 57.2% being di-agnosed with asthma before 16 years of age and 63.7% cases were atopic. For MAAS, genotypes for 1,025 children were available for analyses. General characteristics of cohorts are represented in Table S1-S4.
Association of IL1RL1 variants with specific asthma phenotypes
For these analyses, we focused on the genomic region 400kb up- and down-stream of IL1RL1 (chr2: 102,527,961-103,368,497), which encompasses all of the previously described asthma signals as well as several additional genes (Figure S1). Analyses of the adult asthma GASP and DAG datasets utilizing 3,048 SNPs spanning this region identified initial associations (P<0.01) with multiple traits in one or both co-horts and/or in the meta-analyses (Table 1 and Table S5). In particular, 44 SNPs showed association with blood eosinophil counts, mainly in the DAG cohort (meta-analysis SNP rs77109149 (SL9A2) (C); beta=-0.16 (SE 0.04), P=3.01E-05). Similarly, 15 SNPs showed association with atopy (rs2041747 (IL1R1) (A) meta-anal-ysis; OR=0.53 (SE=0.12), P=7.8E-04). Additional association signals were also observed for childhood on-set asthma (rs12712164 (SLC9A2) (C); OR=0.77 (SE 0.06), P=9.81E-04) in meta-analysis, FEV1 (rs4142132 (IL1RL1) (A); beta=-0.07 (SE 0.02), P=2.67E-05) in meta-analysis and FEV1/FVC (rs113238379 (SLC9A2) (G); beta=0.05 (SE 0.01) P=2.13E-04) (Table S5). Association testing in the childhood cohort (MAAS) for lung function and atopy phenotypes evaluated 2,206 SNPs for association and identified signals meeting ini-tial criteria (P<0.01) for FEV1 (SNP rs12712158 (3’ of SLC9A4, proxy for rs1523204 (r2=0.6) (C); beta=0.06 (SE
0.015) P=1.45E-04), FEV1/FVC (SNP rs1420091 (C); beta=1.27 (SE 0.34) P=1.79E-04) and atopy (rs6543124 (IL18R1) (T); OR=0.65 (CI 0.52-.81) P=1.09E-04) (Table S6-S8).
Table 1. Association results of IL1RL1 variation previously reported to be associated with asthma or asthma related phenotypes in the GASP/DAG cohort.
Results are organised based on significant association to a particular phenotype. Statistically significant results are
highlighted in bold (P<0.05). Linkage Disequilibrium was defined as r2≥0.8.AoO, Age of Onset; LD, Linkage
Disequilib-rium; OR, odds ratio; SE, standard error. *NC = non-coded allele, C = coded allele, †For asthma <16y of age results are represented as OR. ‡rs1420101, rs10192157 and rs13431828 are three of the six tagging variants identified in Stages 1-3 and selected for stratification of functional work.
To complement the genotype based association analyses for known IL1RL1 region polymorphisms, we undertook two sequencing approaches to identify novel, potentially rare, genetic variation. By rese-quencing the exons of IL1RL1 in 94 asthma patients using Sanger serese-quencing we identified 56 variants (Table S9) with 8 resulting in amino acid changes within IL1RL1. We observed 56 SNPs, of which 40 were common and 16 rare (Minor Allele Frequency [MAF]<0.5) in our adult asthma population (N=94). Using a complimentary approach that also addresses intronic and intergenic regions, we enriched the entire chromosome 2q12 locus for region based sequencing using DNA from 200 severe asthma subjects and 200 non-asthmatic, non-atopic control subjects. These analyses identified 4,107 variants spanning the region (Table S10); including 36 variants in the coding region of genes (of which 1 was novel) with nine coding region variants in the IL1RL1 gene (Table S10). Additional variation including structural variation e.g. insertion/deletions spanning the region were also identified (Table S10). Using this novel sequencing data generated in the case/control cohort we completed an association analysis for severe asthma as an outcome using sequencing allele counts. In these analyses there were 4,107 variants (SNPs, indels etc.) for analyses in the 200 cases and 200 controls. Using the same initial screening of association P<0.01 we identified 71 variants meeting criteria (Table S10). The most significant variant, rs34504747 (IL1R1), generates an insertion (C/CCTT), case frequency 0.0004/control frequency 0.1093, P=1.87E-11 (Table S10).
Comparison with previously reported IL1RL1 asthma risk SNPs and SNP prioritization for functional studies
In our Stage 1 analyses we identified 32 SNPs associated with one or more of the asthma related traits across the association analyses using the initial P<0.01 screen (GASP, DAG, GASP/DAG, MAAS, re-se-quencing) (Table S5-S8). To prioritize SNPs for further functional investigation, we also compared our results to the published literature and thus identified32 SNPs associated with asthma or asthma related
phenotypes (Table S11). To provide initial comparison between our findings and the published SNPs we ex-amined signals for these SNPs in the adult DAG/GASP cohorts as these represented the largest datasets (Ta-ble 1). This showed that 11 of the 32 SNPs we identified in the Stage 1 asthma-subtype analysis (Ta(Ta-ble S5-S8) had previously been associated with asthma or related traits (Table 1). Interestingly, multiple SNPs showed excellent concordance with previous reports, e.g. the rs1420101A allele has previously been associated with elevated blood eosinophil counts19 and this association was also present in our cohort. As anticipated, the
blood eosinophil counts and age of asthma onset phenotypes showed the greatest number of associated variants as previously identified for this locus.19,20
Combining published signals with the associations from Stage 1, we identified multiple blocks of associa-tion based on SNP LD r2>0.8 (Figure S2, data not shown). Next, a single tagging independent SNP (r2<0.3
with other signals), from each of the LD blocks, was selected for further investigation, giving an initial panel of 10 SNPs (Stage 2) to be used for stratification of functional studies (Figure S3, Table S12). Of these SNPs, three presented with minor allele frequencies that did not allow for a reasonable call rate in our cell populations and were therefore not amenable to prospective in vitro work (MAF<0.10). An additional SNP (rs1476999) was considered low priority and excluded due to the limited association to this LD block (2 SNPs). Therefore, six priority signals were taken forward for further functional analysis tagged by 6 SNPs (Stage 3, Table 2). These SNPs span the entire genomic region and were in low LD based on r2<0.3
(Figure 2). Given its functional implications we also included the TIR domain IL1RL1 coding region varia-tion to specifically understand the role of this series of amino acid changes in the ST2 protein tagged by rs10192157, even though rs10192157 and rs1420101 showed a moderate level of LD (r2=0.34).
Table 2. Summary of six tagging variants identified in Stages 1-3 association analyses and selected for stratifica-tion of funcstratifica-tional work.
MAF, minor allele frequency; LD, linkage disequilibrium. *NC = non-coded allele, C = coded allele
†rs13431828 (C),7–9 rs1420101 (A)5,10 and rs10192157 (C)8,9 were previously reported as asthma associated SNPs in the
literature, ‡Childhood asthma defined as age of onset <16 years, §Meta-analysis was performed in GASP and DAG, ^rs10192157 (Thr549Ile) tags a haplotype of variants leading to several amino acid changes in the IL1RL1 coding region, rs4988956 (Ala433Thr), rs4988957 (Asn455Asn), rs10192036 (Gln501Lys), rs10204137 (Gln501Arg), rs4988958 (Ser525S-er), rs10192157 (Thr549Ile), rs10206753 (Leu551Ser).
Figure 2. Linkage disequilibrium map of the six variants identified in Stages 1-3 and selected as tagging SNPs for
func-tional study. Figure identifies the level of LD between signals identified based on r2 values. Image generated using
the EUR population of the Phase I cohort of the 1000 genomes study via the LDmatrix tab of the online software tool LDlink 3.0, available at https://ldlink.nci.nih.gov/.
Priority SNPs act as eQTL for both membrane and soluble ST2 encoding transcripts in lung tissue
Lung eQTL analyses identified that 4 of the 6 priority SNPs (rs13431828, rs1420101, rs10192157 and rs990171) act as eQTLs for IL1RL1 in whole lung tissue (Table 3). The rs13431828 asthma risk allele (T) was associat-ed with a decrease in combinassociat-ed (1 probeset; X100148210_TGI_at), and separately, soluble (2 probesets; X100148162_TGI_at and X100312840_TGI_at) IL1RL1 mRNA expression (IL1RL1-a). The rs1420101 (A) and rs990171 (C, lower lung function) risk alleles showed similar associations with the combined and the sol-uble-isoform specific probesets, and in addition showed association with transmembrane (1 probeset; X100302151_TGI_at) IL1RL1 mRNA expression (IL1RL1-b). Effects sizes were 10 fold greater for the com-bined/soluble isoform compared to the membrane variant. The asthma risk (C) allele of the (coding- variant tagging) SNP rs10192157 was associated with a decrease in combined and soluble IL1RL1 mRNA expression (in 2/3 probes). In summary, these data show that four of the priority SNPs are eQTL in lung tissue and all four trait risk alleles are associated with reduced levels of IL1RL1 mRNA in lung tissue, with the strongest effect on the soluble IL1RL1 isoform (Table 3).
Table 3. Association between priority SNPs with both membrane and soluble IL1RL1 mRNA levels in lung tissue.
A Bonferonni corrected P-value of P<2.0E-04 was considered significant, with statistically significant results are high-lighted in bold. MAF, Minor allele frequencies are reported in the EUR population. *NC = non-coded allele, C = coded allele. †Proxy for rs4141632. ‡Proxy for rs12465392.
IL1RL1 SNPs were not identified as eQTLs in airway wall biopsies in healthy subjects
Examination of the six SNPs for IL1RL1 eQTLs in bronchial brushings as well as in bronchial biopsies of healthy volunteers did not identify any eQTL for the six priority SNPs (Table S13).
Effect of IL1RL1 SNPs on baseline IL1RL1 expression in bronchial epithelial cells of asthma patients and healthy controls
To determine the effect of chromosome 2 SNPs on IL1RL1 expression in airway structural cells, we ex-amined the effect of SNPs on baseline expression of IL1RL1 mRNA isoforms and soluble ST2 (IL1RL1-a) protein in samples from cultured human bronchial epithelial cells (HBECs) isolated from both control and asthma patient donors. We combined cases/controls to maximise power, and for significant eQTLs in the combined dataset we additionaly examined if the association of the SNP with IL1RL1 expression was present in the asthma samples separately (n=44).
We identified that the IL1RL1 intronic variant rs1420101 had an effect on the mRNA levels of all three
1RL1splice forms in HBECs cultured in vitro in keeping with a significant role for this SNP in regulating IL-1RL1 expression in the lung. The presence of the asthma risk allele (A) resulted in lower levels of total,
sol-uble and membrane transcripts of IL1RL1 (P<0.05) (Figure 3 Panels A, B, C). These results were confirmed at the protein level, where levels of soluble ST2 in HBEC supernatants were lower in the presence of the asthma risk allele (A) (P<0.01) (Figure 3 Panel D). We also looked for these observations in cells isolated from asthma patients and were able to observe the same trend within these asthma only samples. SNP
rs13431828 was associated with IL1RL1 mRNA levels encoding the transmembrane, but not with those encoding the short isoform in HBECs, with the presence of the asthma associated allele (T) resulting in higher levels of full-length IL1RL1 mRNA (P<0.05) (Figure 3, Panel E and Figure S4). At the protein level, the risk allele was associated with elevated soluble ST2 levels in cellular supernatants (Figure 3, Panel H). This was also observed in cells isolated from asthma patients only.
For rs4141632, the (severe asthma) risk (G) allele resulted in a higher level of total IL1RL1 mRNA with no effect on transmembrane or soluble protein encoding mRNA isoforms or protein levels (Figure 3, Figure S4 and S5). For rs12465392 (SLC9A2, intronic) SNP, the asthma risk allele (A) was associated with a lower level of total IL1RL1 mRNA (Figure 3), an effect that was also evident in the asthma samples separately.
Figure 3. Baseline IL1RL1 mRNA and soluble ST2 protein levels are driven by SNPs in cultured human bronchial epithe-lial cells. In cultured HBECs an decrease in IL1RL1 expression for all isoforms at the mRNA level and for protein levels of soluble ST2 in cellular supernatants can be observed in the presence of the asthma risk allele (A) for rs1420101 (Panels A-D, P<0.05). Increased expression of membrane IL1RL1 was also observed in carriers of the asthma risk allele (T) for rs13431828 (Panel E, P=0.0375), and for total IL1RL1 in carriers of the severe asthma risk allele (G) for rs4141632 (Panel F, P=0.024) and for carriers of the GG genotype (associated with improved lung function) for rs12465392 (Panel G, P=0.025). An increase in soluble ST2 levels in supernatant could also be observed for rs13431828, where carriers of the asthma risk allele (T) presented with an elevated level of protein (Panel H, P=0.048).
Figure 4. Asthma relevant microenvironments modulate IL1RL1 mRNA levels and soluble ST2 protein levels in bronchi-al epithelibronchi-al cells isolated from asthma patients and cultured in vitro. Stimulation of cells with 50µg/ml House Dust Mite (HDM) for 24 hours resulted in increased release of soluble ST2 into the cellular supernatant (Panel A, P=0.01, n=18). RV-16 (MOI:1) stimulation for 24 hours did not significantly influence ST2 protein release in the cell supernatants (Panel B, P=0.50 n=18). HDM stimulation resulted in a 3.5-fold reduction of membrane IL1RL1 mRNA (Panel C, P<0.05, n=15), while stimulation with RV16 (MOI:1) for 24 hours reduced soluble IL1RL1 mRNA levels 4.4-fold (Panel D, P<0.05, n=15). IL-33 stimulation did not influence IL1RL1 mRNA levels, however did induce IL8 mRNA demonstrating cell acti-vation (data not shown).
Effect of asthma relevant stimuli on IL1RL1 expression
Next, we considered the possibility that disease state and/or relevant micro-environmental triggers may regulate IL1RL1 expression, in a SNP dependent fashion. Therefore, we investigated the effect of asthma relevant stimulations of cultured HBECs obtained from asthma subjects, on IL1RL1 expression in the pres-ence and abspres-ence of SNPs, i.e. inducible eQTLs. Prior to stratification, we observed no change in soluble ST2 protein levels following stimulation with either RV16 or with IL33 (P>0.05) (Figure 4, Panel A and B), however stimulation of HBECs with house dust mite (HDM) for 24hrs resulted in a 7-fold increase in sol-uble ST2 protein levels in cell supernatants (P<0.01) (Figure 4, Panel A). When considering mRNA levels, in the total population of asthma HBEC cultures, irrespective of IL1RL1 genotype, stimulation with HDM reduced transmembrane IL1RL1 encoding mRNA expression 3.5-fold (Figure 4, Panel C, P<0.05), while RV16 stimulation reduced soluble ST2 expression 4.5-fold (Figure 4, Panel B P<0.05). No alterations in ST2 mRNA were observed in response to IL33.
IL1RL1 variation has an impact on IL1RL1 regulation in response to asthma relevant stimuli
Stratification based on priority SNPs unmasked effects on IL1RL1 expression by IL33, RV16 and HDM in specific genotype carriers (Figure S6 and S7). In particular, rs13431828 and rs1420101, which showed ef-fects on baseline IL1RL1 mRNA and protein expression in HBECs, also influenced IL1RL1 levels in the pres-ence of stimulation. The rs13431828 asthma risk allele (T) was associated with an increase in total IL1RL1 mRNA in response to HDM (2.8 fold) and a decrease in the level of membrane IL1RL1 mRNA (IL1RL1-b) in response to RV-16 (4.8 fold), while these effects were absent in cells homozygous for the CC genotype (Figure 5 Panels A&B). Homozygote carriers of the C allele for rs13431828 (CC) were associated with an increase in release of soluble ST2 in response to HDM, in contrast to T allele (CT/TT) carriers (2.3 fold vs. control, Figure 5 Panel C). However, it is important to note that the baseline levels of soluble ST2 protein in CC carriers were significantly lower that CT/TT genotype carriers and the supernatant ST2 levels post stimulation were not significantly different in different genotype cells.
Cells carrying the AG/GG genotypes (protective) for rs1420101 presented with an increase in total IL1RL1 mRNA following IL33 stimulation (12.9 fold vs. control), however this induction was not present in AA genotype cells (Figure 5, Panel D). This observation was driven by significantly lower baseline levels of total IL1RL1 mRNA in AG/GG genotype carriers, with mRNA levels of total IL1RL1 following IL33 stimula-tion not differing across genotypes. IL33 and HDM stimulastimula-tions led to a decrease in membrane IL1RL1 expression only in the presence of the AA genotype, with A being the asthma risk allele. Additionally, a reduction of membrane IL1RL1 mRNA expression following HDM stimulation was only observed in in carriers of the AG/GG genotypes for rs12465392 (Figure 5 Panel F, P<0.05).
Bioinformatic analyses of SNPs using ENCODE data
To provide initial mechanistic understanding of priority SNPs and SNPs LD blocks we investigated each SNP and its haplotype as defined by LD r2>0.8 for potential functional effects using the ENCODE resource
via HaploReg (Table S14). The largest LD block was identified for SNP rs990171 (n=79 SNPs) and the small-est LD block for SNP rs4141632 (n=11 SNPs). All the invsmall-estigated LD blocks were found to have multiple SNPs positioned in enhancer histone mark sites, DNase hypersensitivity sites, a region where protein binding consensus sequences exist or affecting regulatory motifs. Both cell-type specific transcription factors such as GATA-2, active in mast cells and basophils21 and positively regulating IL1RL1 expression22,
and FOXA1/2, active in bronchial epithelial cells.23 Similarly, two eQTL SNPs (rs10192157 and rs990171)
al-tered Gfi1 transcription factors linked to type 2 inflammation and a regulator of IL1RL1 expression.24 There
was also generic activation-induced transcription factors such as Fos/Jun (AP-1), NF-Kβ and cEBP/p300 found to be bound to these motifs. This highlights potential functional effects for our 6 priority SNPs on both cell-type specific as well as ubiquitous regulation of IL1RL1 expression (Table S14).
Figure 5. SNPs regulate IL1RL1 mRNA and protein expression levels in response to asthma relevant micro-environ-ments. Increased levels of total IL1RL1 mRNA in response to HDM is present only in carriers of the rs13431828 (T) (asth-ma risk allele) (Panel A, P<0.05). In rs13431828 (T) allele carriers there is a decrease in membrane IL1RL1 mRNA in re-sponse to RV-16 in contrast to C allele carriers (Panel B, P<0.05). Increased release of ST2 protein in rere-sponse to HDM was only present in rs13431828 (C) (protective) genotype carriers (Panel C, P<0.01). IL33 induces expression of total IL1RL1 mRNA in carriers of the rs1420101 (G) allele, with A being the asthma risk allele (Panel D), whereas both IL33 and HDM treatment reduce membrane IL1RL1 mRNA in rs1420101(A) risk genotype carriers (Panel E, P<0.05). Cells carrying the rs12465392 (G) allele (higher FEV1) show a decrease in membrane IL1RL1 mRNA in response to HDM (Panel F).
Amino acid changes driven by variants in the ST2 coding region are functional
We next tested functional effects of the membrane IL1RL1 exon 11 haplotypes that are tagged by rs10192157 (Table 4). These haplotypes encode IL1RL1 proteins that present with a 4 amino acid change in the intracel-lular TIR signalling domain.12 The potential functional effects included: 1) IL1RL1-b coding region variants
determine the magnitude of signalling response downstream of IL3325 and influence Toll-like receptor 2 and
4 activation15,16 and 2) IL1RL1 haplotypes determine the anti-inflammatory effects of anti-IL33 and
anti-IL-1RL1 monoclonal antibodies.25 A reductionist recombinant cell line model with a fixed genetic background
was used to facilitate these analyses. HEK-Blue-SEAP cells transfected with empty vector or the two variant
IL1RL1 mRNA encoding the alternative TIR domain ST2 proteins demonstrated the same ability to signal via
NF-Kβ following TNF-α stimulation (Figure 6, Panel A). Escalating doses of recombinant IL33 were able to induce NF-Kβ signalling, in a dose dependent manner, in the two cells lines containing the ST2 protein. In contrast, the empty vector control cell line gave a background signal (Figure 6, Panel B). Cells carrying the asthma risk haplotype (Ala433/Gln501/Thr549/Leu551) tagged by priority SNP rs10192157 (CC) demonstrat-ed a 2.9 fold induction at highest dose of IL33 (50ng/ml) in this cell system which was significantly higher than the modest activity observed for the asthma protective TIR domain haplotype protein (Thr433/Arg501/ Ile549/Ser551) (1.3 fold) (Figure 6, Panel B). Additionally, cells carrying the asthma risk haplotype and stim-ulated with 50ng/ml IL33 were more amenable to the anti-inflammatory effects of blocking either ST2 and IL33 using monoclonal antibodies compared to the alternative haplotype were blocking antibodies had a minimal effect on reducing NF-Kβ signalling (Figure 6, Panel C and D).
Figure 6. Functional analyses of the IL1RL1 TIR risk haplotype in an in vitro reductionist model identifies an exagerated response to IL33 that is more amenable to anti-IL33/ST2 treatment. Transient transfection of HEK-NF-Kβ-SEAP reporter cells with IL1RL1 containing the the two TIR domain polymorphism haplotypes provides a platform to identify differen-tial NF-Kβ signalling. Cells transfected with empty vector, IL1RL1 containing the asthma risk haplotype (Ala433/Gln501/ Thr549/Leu551) or IL1RL1 containing the protective haplotype (Thr433/Arg501/Ile549/Ser551) have the same capacity to signal via the NF-Kβ pathways in response to 10ng/ml TNF-α (Panel A). The presence of the IL1RL1 receptor carrying the asthma risk haplotype identified a 2-fold and 3-fold increase in signalling on stimulation with 10ng/ml and 50ng/ ml of human recombinant IL33 respectively, whereas an attenuated response was observed in the common haplotype (Panel B). The response induced by 50ng/ml IL33 in the risk haplotype was amenable to blocking using either 10µg/ ml anti-IL33 or anti-ST2 leading to an anti-inflammatory effect (Panel C). Whereas effect of blocking IL33 induced in-flammation by anti-IL33 or anti-ST2 was minimal in carriers of the protective TIR domain haplotype (Panel D). *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. N=3 for all experiments.
Amino acid changes driven by variants in the ST2 coding region do not influence TLR signalling
For the TLR series of experiments, we used ILR1L1 overexpression vectors in transfected cells whose func-tionality was confirmed by the capacity to induce ERK1/2 activation after IL33 exposure (Figure S7). We did not observe an effect of overexpression of the two different IL1RL1-b haplotypes on the sensitivity of HEK-Blue cells to TLR2 induced NF-Kβ activity after stimulation using a dilution series of Pam3Cys (Figure S8 Panel A). We also did not observe an IL1RL1-b variant driven effect on TLR4 induced NF-Kβ activity in HEK-Blue cells after stimulation with a dilution series of LPS (Figure S8 Panel B). These studies show that the exon11 haplotype regulates sensitivity to IL-33, while the proposed regulatory function of IL1RL1-b on TLR2 or TLR4 signalling15,16 could not be confirmed.
Discussion
Main findings
In the last decade we have made significant advances in our understanding of the genetic basis of asth-ma, however the translation of these findings to mechanistic insight and information that can be used to identify new drug targets or tailor therapies to asthma patients remains to be realised. We set out to ex-tend our understanding of the IL1RL1 locus in asthma, one of the most reproducible association signals to date, with a particular focus to the potential contribution of the IL1RL1 gene. We provide new insight into the nature of the genetic association with clinical and immunological features of asthma and the mechanistic underpinnings of these associations with respect to IL1RL1 expression and activity. In this unique multiple cohort study, we extend a priori evidence that genetic variation in the IL1RL1 region is important for asthma susceptibility, blood eosinophil counts, age of onset of asthma and atopy in particular. Functional analyses demonstrated that four (rs13431828, rs1420101, rs10192157, rs990171) of our six priority SNPs are eQTLs for
IL1RL1 in lung tissue. These regulatory roles were retained in cultured primary HBECs at both the mRNA
and protein levels, however in some instances for alternative alleles. Interestingly, there was a lack of eQTL effects observed in bronchial biopsies and bronchial brushes taken from controls without respiratory dis-eases, whereas we did observe eQTLs in cultured primary HBECs of asthma patients. We next approached whether the priority SNPs also act as context dependent, i.e. inducible eQTLs, through functional studies. We identified that several of our selected SNPs were able to modulate the effect of the micro-environment, e.g. responses to RV16 or HDM, on IL1RL1 expression in HBECs. These data provide a potential link between
IL1RL1 genetic variants and IL1RL1 regulation of IL33 inflammation in virus-induced exacerbation in asthma,
where the IL33/IL1RL1 axis has been shown to be important.26,27 Finally, we also investigated IL1RL1 coding
region variation and established that cells carrying the IL1RL1 TIR domain asthma risk haplotype presented with an exaggerated inflammatory response to IL33 that is more amenable to the anti-inflammatory effects of either anti-IL33 or anti-IL1RL1 monoclonal antibodies. This has implications for the targeting of IL33/IL1RL1 axis inhibitors to a sub-set of patients likely to gain the greatest clinical benefit and is highly relevant with multiple pharmaceutical companies developing anti-IL33/IL1RL1 approaches for asthma.
Genetic associations
The locus on chromosome 2q12, which includes the IL1RL1 gene, has shown significant replicated asso-ciated with asthma5,7,19,28,29, or asthma-relevant phenotypes such as childhood asthma20,30,31, childhood
asthma with exacerbation30, severe asthma32, asthma with hay fever33, T2 inflammation in asthma34 and
blood eosinophil counts.19 Sentinel SNPs identified in these association studies span the 2q12 region
cov-ering multiple genes, with evidence pointing to multiple, potentially independent, association signals based on linkage disequilibrium.19
In order to characterise further the contribution of genetic variants to features of asthma, we carried out association testing across the GASP/DAG/MAAS cohorts. This identified that chromosome 2q12 region variation is associated with blood eosinophil numbers (key SNP rs77109149), atopy (rs2041747), and childhood onset asthma (rs12712164), with some further evidence for association with lung function (FEV1, FEV1/FVC) (rs990171, rs12465392). We also investigated SNPs previously associated with asthma or asthma related traits in the region, with 11 of these SNPs being associated with asthma related traits in our GASP/DAG analysis, e.g. rs1420101 (IL1RL1, blood eosinophil numbers, childhood onset asthma, atopy), rs13431828 (IL1RL1, childhood onset asthma, atopy), and rs10192157 (IL1RL1, increased blood eo-sinophil levels, childhood onset asthma, atopy), the latter of which tags a series of amino acid variants in
identified 4,107 variants spanning the region, of which 71 met criteria for association with severe asthma including key variant rs34504747 (C/CCTT (IL1R1)). These findings need further replication, however rep-resent the first use of next generation sequencing to investigate this region. Therefore, based on these analyses we selected 6 priority SNPs that identify key signals of association spanning the region for func-tional analyses, three of which were novel associations (rs4141632 (IL1R1), rs990171 (SLC9A4), rs12465392 (SLC9A4)). Overall these data complement and extend the accumulating data suggesting a role for the IL33/ST2 axis that is genetically determined in T2 driven inflammation34,35, severe childhood asthma36,
eosinophilic asthma19 and allergic sensitization.37 eQTLs
Previous work has established that genetic variation in the IL1RL1 locus can act as eQTL for IL1RL1 mRNA, as methylation QTLs in white blood cells and as pQTL for serum levels of sST2.13,25,31,35 In this study , we
specif-ically investigated the functional effects of our 6 priority SNPs in (i) lung tissue, (ii) airway epithelial brush samples and (iii) cultured HBECs at baseline and in the presence of asthma relevant stimuli. Lung tissue eQTL identified 4 of the 6 priority SNPs as eQTLs for IL1RL1, with differential SNP effects observed for mem-brane and soluble isoforms. Interestingly, the presence of the asthma risk allele for all of the selected IL1RL1 SNPs were associated with reduced expression of IL1RL1 mRNA transcripts. The effect size of the associa-tions was greatest for the soluble protein isoform mRNA suggesting that IL1RL1 eQTL effects are more likely to translate to functional effects through regulation of soluble rather than membrane ST2 levels, with the exception of TIR domain effecting rs10192157. This latter SNP has also been identified as an eQTL for IL1RL1 in non-structural cells.38 Our data are in good agreement with the GTEX database and previous work
ex-amining genetic variants that are associated with soluble ST2 mRNA34 and serum levels of soluble ST2.25 Importantly, our data extend this work by offering a comprehensive analysis of all IL1RL1 transcripts (total, membrane and soluble ST2 encoding mRNA) and assessing key genetic variation in the wider gene region. This extends the recently described concept that asthma risk alleles essentially lead to a decrease in soluble ST2 and this lack of decoy receptor diminishes the ability to mitigate the effects of IL33.34
We did not observe an eQTL effect of our priority SNPs in the biopsy and bronchial brush datasets gen-erated from tissue samples obtained from volunteers without respiratory diseases, which is in contrast to a recent study that identified 3 IL1RL1 SNPs (rs12712135 )tagged by rs1420101), rs1041973 (tagged by rs13431828), rs10185897 (tagged by rs13431828)) showing association with membrane IL1RL1 mRNA in bronchial brush samples from asthma patients.35 This can potentially be explained due to differences in
sample cell composition. To provide greater insight, we therefore completed reductionist eQTL analyses in cultured primary bronchial epithelial cells (HBECs) isolated from controls without respiratory diseas-es and asthma patients. Here we identified that rs1420101 data complemented that in our lung tissue data and a recent report34, with the asthma risk allele being associated with decreased total, membrane
and soluble-encoding mRNA levels and soluble protein levels. Interestingly, we demonstrated that the rs13431828 asthma risk allele (T) was associated with a higher levels of membrane and soluble mRNA and soluble ST2 protein, i.e. in the opposite direction to that observed in the lung tissue analyses. These data provide an additional indication that some of the IL1RL1 SNPs act as eQTL in a tissue and cell type specif-ic manner, potentially with inflammatory cells contributing to the lung tissue findings, e.g. mast cells, basophils, Th2 cells, ILC2s and eosinophils. Therefore, cell-type specific eQTL datasets in asthma are ur-gently needed, however our new data suggest asthma risk alleles across the different signals particularly associate with lower levels of soluble IL1RL1-a in the lung which may attenuate the ability of soluble ST2 to mitigate IL33 activity. The loss of a regulatory loop to counteract the effects in genetically susceptible individuals may at least explain the increased risk of blood eosinophilia, childhood asthma and atopy, all outcomes that IL33 may play a driving role in.
Inducible eQTLs
We tested for inducible eQTLs of our priority SNPs by culturing HBECs in the presence and absence of RV-16, HDM, and human recombinant IL33. In general, RV-16 stimulated HBECs showed a decrease in soluble ST2 mRNA, however SNP-specific effects were observed on stratification for rs13431828, where RV-16 decreased membrane IL1RL1 mRNA only in carriers of the protective allele (C/T and TT). Interest-ingly, rs13431828 is in moderate LD (r2=0.44) with rs1558641, a SNP that has been associated with asthma
exacerbations in children, with ~80% of exacerbations driven by Rhinovirus.30 RV16 has been previously
associated with the IL33/IL1RL1 axis, with HBEC stimulations with RV16 inducing IL33 levels in cell su-pernatants57 and IL33 levels in bronchial mucosal lining fluid correlating with the severity of RV16 in-duced exacerbation in humans.26 These data fit well with our observation that this variant is a risk factor for
childhood atopic asthma where virus induced effects are likely to be most prominent. Taken together these data suggest that in carriers of the rs13431828 asthma risk allele (C) there is inhibition of the RV16 driven attenuation of IL1RL1 receptor levels in HBECs, making these cells more sensitive to the IL33 released as a consequence of the viral infection. This increased sensitivity might in part may contribute to RV16 induced asthma exacerbations.
When considering the aeroallergen HDM, another relevant environmental agent involved in allergic asthma, our data identified that HDM-exposed HBECs display reduced expression of the IL1RL1 transcript encoding the transmembrane receptor. HDM driven effects on IL1RL1 in HBECs however appear to be SNP dependent. While carriers of the protective allele for rs13431828 (CT/TT) responded to HDM by an increase in total IL1RL1 mRNA, it was homozygote carriers of the asthma risk allele (AA) for rs1420101 that presented with attenuated levels of the IL1RL1 transcript encoding the transmembrane receptor follow-ing HDM stimulation. A similar response was observed followfollow-ing stimulation with IL33, where presence of the asthma risk allele (A) for rs1420101 was again associated with attenuated IL1RL1 mRNA expression. Considering our findings in relation to IL1RL1 inducible eQTLs, we present the signals tagged by rs1420101 and rs13431828 as regions of particular importance for IL1RL1 regulation both at baseline and in response to these allergic triggers. The LD block tagged by rs13431828 includes risk alleles for childhood asthma, atopy in our analyses and asthma exacerbation30, whereas rs1420101 is associated with blood
eosino-phils, childhood asthma and atopy, Type-2 inflammation and asthma19,34, all relevant to allergic disease
potentially driven by allergens such as HDM.
Interestingly, ENCODE data indicated that rs1420101 and rs13431828 share an overlap in the transcription factor binding motifs which are disrupted by the polymorphisms present in these LD blocks, such as GATA2, which regulates IL1RL1 expression in mast cells and basophils22 and has been shown to mediate the effects
of HDM in the human lung has been reported.39 Of note also is that several of the eQTL SNPs resulted in
changes in type 2 inflammatory transcription factors that are known to regulate IL1RL1, e.g. Gfi1. However, a large range of ubiquitously expressed transcription factors, were likewise modified by changes driven by rs13431828 and other eQTL SNPs, indicating that the effect on IL1RL1 expression by different stimuli may be driven through different transcription factor binding motif interactions. As in the case of the signal tagged by rs13431828, these effects may be cell type specific (Table 4). This cell specific eQTL is further supported by the different response observed at baseline in cultured HBECs vs HBECs taken straight from brushings (which may be under the direct influence of local inflammatory processers) and lung biopsy samples, which are complex tissues consisting of a variety of structural and immune cells. While we present a comprehen-sive analysis of functional effects of IL1RL1 region SNPs for IL1RL1 gene expression and function, we acknowl-edge that functional effects on other genes (i.e. eQTL effects on e.g. IL18R1) may be of relevance as well, yet these were beyond the focus of our current investigations to advance our understanding of the contribution of genetic variants to IL1RL1 biology in the context of asthma.
Table 4. Summary table identifying functional effects of priority tagging SNPs on IL1RL1 expression and activity.
Childhood asthma defined as age of onset <16 years.
†Identifies effect of variation on soluble ST2 protein expression in HBEC supernatants. ‡Effect driven by alterations in baseline expression with no difference between genotypes in stimulated environment.
Coding variants
Finally, we examined the functional effects of the rs10192157 SNP that is in complete LD with several polymorphisms encoding 4 amino acid changes; Ala433/Gln501/Thr549/Leu551 being the asthma risk haplotype and Thr433/Arg501/Ile549/Ser551 being the protective haplotype. In agreement with other reports, we show that the asthma risk haplotype leads to enhanced NF-kb activity after IL33 mediated
IL1RL1 signalling in a reductionist cell model.25,29 It has been previously suggested that IL1RL1-b may also
affect TLR2 and TLR4 signalling16, but we were unable to show supporting evidence for this in an in vitro
TLR2 and TLR4 reporter cell models. However, we do show also cells carrying the asthma risk ST2 protein are more amenable to the anti-inflammatory effects of anti-ST2 and anti-IL33, which has important ther-apeutic implications for potential stratified medicine approaches, especially in light of current pharma-ceutical development of an IL33/IL1RL1 antagonist for use in asthma.
Conclusion
This study has significantly advanced our understanding of both the phenotypic and the functional ef-fects of polymorphisms in the IL1RL1 locus in the context of asthma. We have confirmed and extended genetic association to specific features of asthma, including severe asthma, blood eosinophil counts, childhood onset asthma and atopy including a SNP selection strategy for functional work. All 6 of the SNPs identified for functional analyses show effects on IL1RL1 regulation, complementing and extend-ing the literature, particularly by examinextend-ing eQTLs in lung tissue and bronchial epithelial cells under
different environments. These data suggest that asthma and asthma-phenotype related risk alleles, at multiple loci, significantly affect IL1RL1 mRNA and protein levels in a tissue and isoform specific way. Overall our study therefore highlights the complexity of this susceptibility locus for asthma and iden-tifies multiple SNP driven mechanisms that contribute to the genetic association signals identified in GWAS. The concept of multiple independent mechanisms contributing to asthma may at least in part ex-plain why this locus represents one of the most reproducible association signals in asthma GWAS to date.
Methods
Cohorts
Genetics of Asthma Severity and Phenotypes (GASP) cohort
Asthmatic individuals aged 16-60 were recruited from across the United Kingdom as part of an Asthma UK initiative to generate a cohort that is enriched for patients with British Thoracic Society Step 3 and above (moderate-severe asthma).40 Subjects had extensive clinical characterisations (Table S1). Asthma
was defined as a doctor’s diagnosis of asthma through the presence of symptoms and medical treatment, while age of onset of asthma was determined through patient records. Asthma related clinical pheno-types used in the current study focussed to lung function (FEV1 pre bronchodilator, FEV1/FVC pre-bron-chodilator), atopic status (positive skin prick test), Blood Eosinophil Count (x109/L) and Blood IgE levels (KU/L). Lung function tests for FEV1 and FVC were carried out through spirometry using a Vitalograph Medisoft BB5500 Whole Body Plethysmography System. Total peripheral blood eosinophil levels were calculated using a counting chamber while total Immunoglobulin E (IgE) levels were measured by Immu-noCAP™. Finally, atopy was defined as a positive response to a skin prick test (SPT) to any allergen from a panel of 4-24 allergens.
Dutch Asthma GWAS (DAG) cohort
The DAG cohort consists of 469 trios (patient – spouse – child) ascertained through a proband with asth-ma, combined with an additional case-control study of 452 asthmatics and 511 controls.17 Of these, we
selected 909 asthma patients who underwent the same, standardized, comprehensive evaluation for asthma at Beatrixoord Hospital, Haren, The Netherlands between 1962-2000. Asthma was defined as a doctor’s diagnosis of asthma, asthma symptoms, and presence of bronchial hyper-responsiveness (BHR). FEV1 was measured using a water-sealed spirometer (Lode Spirograph type DL, Lode b.v., Groningen, The Netherlands). Total peripheral blood eosinophils were counted in a counting chamber and IgE levels were measured in serum by an enzyme-linked fluorescence assay (Mini Vidas, Biomerieux Inc., Marcy, France). In subjects older than 12 years, intracutaneous tests with 16 common aeroallergens were per-formed. In children younger than 12 years, a skin prick test was performed with 10 allergens. Subjects with a positive response to one or more intracutaneous or skin prick tests were considered to be atopic. Age of asthma onset was based on data from medical records and questionnaires, indicating the start of asthma symptoms (Table S1).
Manchester Asthma and Allergy Study (MAAS) cohort
MAAS is a birth cohort study for which subjects have been randomised prenatally, and which follows the development of both asthma and other atopic disorders in a cohort of infants at high risk of atopy. Subjects were recruited in the maternity catchment area of Wythenshawe and Stepping Hill Hospitals, comprising 50 square miles of South Manchester and Cheshire, United Kingdom (Table S3).
Lung Tissue Cohort
Non-tumor containing lung tissue was collected from patients who underwent lung resection surgery at three participating sites: Laval University (Quebec, Canada), University of British Columbia (Vancouver, Canada) and University Medical Center Groningen (Groningen, The Netherlands) and cohort characteris-tics have been described previously.41
Normal Values of Inflammatory Variables From Healthy Subjects (NORM) cohort
77 respiratory healthy subjects were derived from the Normal Values of Inflammatory Variables From Healthy Subjects (NORM) study (NCT00848406).42 Current smokers and never smokers older than 18
years were recruited into this study. Subjects were considered respiratory healthy if they had no respi-ratory symptoms, no history of respirespi-ratory disease and normal pulmonary function. Normal pulmonary function was defined as a post bronchodilator FEV1/FVC higher than lower limit of normal, absence of AHR to methacholine (PC20 <16mg/mL) and absence of reversibility (FEV1 %predicted to Salbutamol < 10%). Subjects were excluded if they used inhaled or oral corticosteroids within the last 5 years, or for a period of 5 years or longer.
The Asthma Human Bronchial Epithelial Cell (AHBEC) cohort
The asthma human bronchial epithelial cell (AHBEC) cohort consists of primary bronchial epithelial cells from 33 individuals with a doctor’s diagnosis of asthma and 18 control subjects recruited from Notting-ham and Leicester. Cells were obtained through endobronchial biopsies carried out as per standard pro-cedure.43,44 These cells were used for all primary human cell culture, as described. Clinical characterization
of these patients included lung function tests for FEV1 and FVC. Total peripheral blood eosinophil levels as well as sputum eosinophil and neutrophil levels were calculated using a counting chamber (Table S4).
Genotyping GASP
Participants in the GASP cohort were genotyped using two platforms, 744 subjects initially genotyped us-ing the Affymetrix Axiom® UK BiLEVE array and 2172 subjects genotyped subsequently usus-ing the related Affymetrix Axiom® UK Biobank array. In each genotyping batch samples were excluded if: (i) their genet-ically inferred gender did not match their reported gender; (ii) they had outlying heterozygosity within the batch (outside either 2 or 3 standard deviations from the mean depending on batch); (iii) they had a call rate < 95% across genotyped variants; (iv) they were cryptically related to another sample, in which case 1 sample of the pair was removed; (v) the sample shows significant deviation from European an-cestry as determined by a plot of the first two principal components. Following quality control, 692,060 SNPs were available following exclusion of monomorphic SNPs and SNPs only found on a single array, for 2,536 subjects, to contribute to these analyses.
DAG
Participants in the DAG cohort were genotyped on two platforms, the Illumina 317 Chip and the Illumi-na 370 Duo Chip (IllumiIllumi-na, San Diego, CA). Quality control (QC) was performed per chip with exclusion of individuals with a missing genotype call rate >0.01, related individuals (identity by descent sharing (IBS) >0.125) and non-Caucasian subjects, as assessed by principal components analysis performed with EIGENSTRAT.45 SNPs were excluded with a missing genotype rate >0.01, a Hardy-Weinberg equilibrium P-value <10-7 and a MAF <0.01. Markers with Mendelian errors in phase I were excluded from analysis. Following quality control, the chips were merged and SNPs not available in both cohorts were excluded from the dataset. A total of 294,775 SNPs remained. Imputation was performed using IMPUTE 2.0 against
the reference data set of the CEU panel from the 1000 Genomes project (version March 2012).46 A subset of
this cohort (n=94) was selected for targeted resequencing of the IL1RL1 gene. We developed 23 primer pairs with primer 3 software spanning the distal promoter region, exon 1a, intron 1a, proximal promotor, exon 1b, intron 1b, exon 2, exon 3, exon 4, exon 5, intron 5, exon 5E, intron 5E, exon 6, exon 7, intron 7, exon 8, intron 8, exon 9, exon 10, intron 10 and exon 11. PCR products were Sanger-sequenced with Applied Biosystems 3730x1 DNA analyser. DNA variants were analysed using Mutation Surveyor Software.
MAAS
For the MAAS cohort, DNA was extracted from a total of 1025 individuals, who were genotyped using the Illumina Human610-Quad array providing 2206 SNPs for analyses post QC in the chromosome 2 region.
Lung Tissue Database
In the Lung Tissue Database DNA samples were genotyped with Illumina Human1M-Duo BeadChip ar-rays, and gene expression profiles were obtained using a custom Rosetta/Merck Human RSTA Custom Affymetrix 2.0 microarray. Gene expression data are available on the Gene Expression Omnibus acces-sion number GSE23546 and platform GPL10379. Imputed SNP data were available for 1,095 of the 1,111 subjects. The final dataset included n=1088 subjects.
Next-Generation DNA Sequencing (NGS)
DNA from 200 severe asthma cases (BTS 4, 5) from GASP and 200 asthmatic, atopic, non-wheeze controls from the Nottingham Gedling cohort,47 were selected for resequencing. Subjects were
matched for age and gender (Table S2). Next-generation Illumina sequencing of the 2q12.1 region was outsourced to Source Bioscience (Nottingham, UK) and was carried out using the SureSelect enrichment approach. The chromosome 2 locus previously associated with asthma5,6 [GRCh37.p9] was the focus and
120 base pair paired-end long read oligonucleotides (baits) were designed using the SureSelect™ e-array design software. Bait tiling (X5) was used across the region, presenting with a capture size range of 500Kb to 1.5Mb. The initial target region was 470,788bp; using 13,606 baits achieved 79.58% coverage of this region. Samples were pooled for sequencing (3 pools for cases and 3 pools for controls). Next-generation sequencing was carried out on these six samples on two separate lanes, one for cases and the other for controls, using the Illumina HiSeq2000™ systems pipeline (San Diego, USA). Sequencing used a paired end design using 100bp reads.
Bronchial brushes and biopsies
In the NORM study bronchial brushings were collected using a Cellebrity brush (Boston Scientific, Mas-sachusetts, USA) for genome wide gene expression profiling using microarrays. The methods for RNA extraction, labeling, microarray processing and Principal Component Analysis (PCA) are described of the bronchial biopsies in the Supplementary Methods.
Selection of genetic region and IL1RL1 SNPs Selection of region
For the phenotypic analyses we selected SNPs with a minor allele frequency (MAF) ≥0.01 located 400kb up- and downstream the IL1RL1 gene (chr2: 102,527,961-103,368,497). There were 3148 and 3048 SNPs with snptest infoscore >0.3 present in the GASP and DAG cohorts, respectively. Annotated SNP location and function was determined with the use of HaploReg v4.1.48 All genetic data were aligned to assembly
Selection of SNPs for cell based analysis
SNPs of interest were selected using the following criteria: (i) prior evidence from the literature (Table 1, Table S11), (ii) significant association with asthma-subtypes in our genetic association analysis (p<0.01), with prioritization of SNPs associated to multiple subtypes (iii) a minor allele frequency over 10% to fa-cilitate subsequent in vitro analysis and (iv) independence based on r2<0.3 (in the 1000 genomes CEU
population.49 Levels of linkage disequilibrium were identified for these SNPs utilising the online software
LDlink50,51 (Figure S2 and S3). From this, SNPs were prioritised based on LD (LD r2<0.3) Figure 2) and
select-ed for further study (Table 2). A tagging SNP was chosen from each key haplotype block/signal of interest after the selection process ultimately give 6 priority SNPs.
Functional characterisation of IL1RL1 variants in primary bronchial epithelial cells Cell culture and collection of RNA and protein lysate
Passage 2 and 3 HBECs obtained from bronchial brushes and biopsies from the AHBEC cohort as previ-ously described52, were cultured on PureCol Type-I Bovine collagen (Advanced BioMatrix, 5005-B) in fresh
growth factor-supplemented medium (BEGM, Lonza). BEGM was changed to basal medium 24 hours pri-or to stimulation with either HDM (50µg/ml) (Greer XPB70D3A25 (Lot: 23187)), human recombinant IL33 (50ng/ml) (Abnova, P3638) or PBS as a vehicle control. For stimulation with RV16 (Public Health England), BEGM was changed to infection medium (BEGM-I), i.e. BEGM lacking Bovine Pituitary Extract, 24 hours prior to infection with a virus MOI of 1. Cells were incubated for 24 hours. Protein and RNA lysates were collected as previously described.53
Quantitative PCR
HBEC complimentary DNA (cDNA) was synthesised from 1µg RNA using Superscript II (Invitrogen, UK) and random hexamer primers according to the manufacturer’s instructions. TaqMan® Quantitative PCR (qPCR) was then utilised to quantify mRNA levels of IL1RL1 and was performed and analysed as previous-ly described.53
Genotyping of primary bronchial epithelial cells
DNA was extracted using the Qiagen QIAamp® DNA Mini and Blood Mini Kit according to the manufac-turer’s instructions. SNP Genotyping was then carried out using TaqMan® Pre-designed assays.
Quantification of soluble ST2 protein levels
ST2 in cellular supernatants was measured using Luminex assays (supplied by R&D, product code LX-SAHM) according to the manufacturer’s recommendations using a custom Magnetic Luminex Screening Assay with a Human Premixed Multi-Analyte Kit (R&D systems). Each experimental supernatant was as-sayed in duplicate.
Functional characterisation of IL1RL1 coding region variants using recombinant expression Preparation of IL1RL1 coding region transfection plasmids
The open reading frame of membrane IL1RL1 containing the exon 11 Asthma protective/risk haplotypes for the TIR domain haplotype was amplified from mRNA to generate two expression cassettes contain-ing either; Ala433-Glu501-Thr549-Leu551 or Thr433-Arg501-Ile549-Ser551 haplotypes. Primers includcontain-ing a consensus Kozak sequence and restriction enzyme sites: 5’primer: 5’-ACTTGCTAGCGCCACCATGGG-GTTTTGGATC-3’ and 3’ primer: 5’-ACTTGCGGCCGCCTATTGCTTCTGG GCAGCC -3’. The PCR products were cloned into the pGEM®-T Vector System II (Promega A1380), subcloned (NheI and NotI) into pCDH-CMV-MCS-EF1-copGFP (Addgene) and sequence verified.
IL1RL1-HEK Reporter cells system
HEK293 cells genetically modified to contain an NF-Kβ Secreted Alkaline Phosphatase Reporter (NovusBio NBP2-26260), were cultured in DMEM containing 4.5 g/L -glucose and 4 mM L-glutamine (Sigma D5796) supplemented with 10% FBS, 1 mM sodium pyruvate, 100 units/ml Penicillin, 100µg/ml Streptomycin and 500µg/ml G418. Cells were transfected with the IL1RL1 plasmids in a 96-well plate format using the Tran-sIT-LT1 lipid transfection reagent (Mirus Bio, MIR2300) as per manufacturer’s instructions. Empty vector (pCDH-CMV-MCS-EF1-copGFP) was used as a transfection control and adjustments were made to correct for differences in plasmid size. Following a 24 hour period, cells were incubated for 1 hour with either 10µg/ ml of either an anti-IL33 (Invivogen, mabg-hil33-3) or anti-ST2 antibody (R&D, MAB523) followed by stimu-lation with a range of concentrations of human recombinant IL33 (ABNOVA). NF-Kβ -activated SEAP release was measured 24 hours following IL33 stimulation, using a commercial SEAP Reporter Assay Kit (Invivogen). 10µl of cellular supernatants were used and activity measured as per manufacturer’s instructions. For the TLR experiments, an analogous system was used, however HEK293-NF-Kβ cells containing recombinant TLR2 or TLR4 prior to IL1RL1 transfection were used.
Western blots
HEK293T cells stably transfected with the protective or the risk haplotype of IL1RL1 (in pcDNA3.1) were seeded into 6 well plates (1x106 cells /well) in DMEM (10%FCS), cultured for 24h, followed by serum
star-vation (0% FCS/DMEM) overnight. Cels were stimulated after o/n serum starstar-vation with IL33 at 30 ng/ ml for 0/10/20/30 minutes in DMEM (0%FCS), followed by aspiration of the media and lysis of the cells using Laemmli buffer (200 uL per well), scraping of the well and collection of the lysate. Lysates were immediately boiled for 10 minutes and stored at -80. Lysates were anayzed using 10% PAA gel (home made by standard TRIS/SDS/Acrylamide/APS/TEMED) followed by transfer to nitrocellulose using meth-anol-containing blotting buffer (200 Amp, 2 hr). Blocking of the nitrocellulose is in 5% ELK/TBS, washing in 1x TBST. Antibodies for detection are from CellSignalling.
Statistics
Genotype-phenotype associations
IL1RL1 region SNP associations in the GASP and DAG cohort with FEV1, FEV1/FVC, blood eosinophils, total
IgE levels, atopy and age of asthma onset were performed using the software package snptest v2.5β54, using
an additive model. Eosinophils and IgE levels were logarithmically transformed before analysis. Age of on-set was analyzed both as a continuous variable and a quantitative trait with cases being defined as having asthma onset <16 years of age (childhood onset asthma).5 FEV1 and FEV1/FVC analyses were corrected for
age, gender and height. Eosinophils, IgE and atopy were corrected for age and gender and the age of asthma onset analysis were corrected for gender. Meta-analysis was performed in METAL55 using a fixed effect
mod-el. P-values of <0.01 were considered statistically significant and of interest as there was a priori evidence of an association between these SNPs and asthma (related traits)5,34,41 and there was strong LD between
multi-ple SNPs, an approach that was published previously.56 Next-Generation Sequencing
For the association testing in the sequencing cohort pooled data generated by next-generation se-quencing was mapped to the NCBI GRCh37 genome template and compared to the NCBI SNP database (dbSNP) to determine known variation in the region. Realignment around insertion deletions and reca-libration of quality scores was carried out using the freeware software programme GATK57, while novel
variant detection was carried out using the freeware software program SyzgyÆ.58 Variants were called
of accuracy of the called SNPs within the analysed region. Variants selected were tested for association with asthma risk using Fisher’s exact tests on allele counts per case/control pool, utilising allele counts produced by the Syzygy.
Lung & Bronchial Biopsy eQTL
For the lung eQTL analyses first, cohort specific (Groningen, Laval and UBC) principal components (PCs) were calculated based on residuals from linear regression models on 2-log transformed IL1RL1 expres-sion levels (of each IL1RL1 probe separately; X100148162_TGI, X100148162_TGI, X100312840_TGI_at and X100302151_TGI_at) adjusted for disease status (alpha-1 antitrypsin deficiency, idiopathic pulmonary fibro-sis, pulmonary hypertension, cystic fibrofibro-sis, other disease), age, gender and smoking status (never/ever/un-known). PCs that explained at least one percent of the total variance were saved and included as covariates in the main analysis; these were 14 PCs for Groningen and Laval, and 16 for UBC. Previously eQTL analysis using the same probes have been described by Akhabir et al.59 Second, in each cohort separately, linear re-gression analysis was used to test for association between SNPs 400 MB flanking of the IL1RL1 gene and 2-log transformed gene expression levels in SNPtest v2.5β.54 SNPs were tested in an additive genetic model and the models were adjusted for disease status, age, gender, smoking status and the cohort specific PCs. Finally, SNP effect estimates of the three cohorts were meta-analyzed using fixed effects inverse variance models. We identified no SNPs under the probe with an allele frequency >10%. A linear model was used correcting for smoking, gender and age and PC explaining more than 1% of the variance where stated. Four probesets were utilized for eQTL analysis; Probeset 1 (X100148210_TGI_at) provided information on the expression of all IL1RL1 transcripts, Probeset 2 (X100148162_TGI_at) and Probeset 3 (X100312840_TGI_at) on transcripts encoding soluble ST2 and a less putative variant (isoform 3), and Probeset 4 (X100302151_TGI_at) on the transmembrane IL1RL1 transcript. For the six priority SNPs a P<0.002 (Bonferroni correction) was consid-ered significant.
For the eQTL analysis of bronchial biopsy and brush gene expression a linear model was used correcting for smoking, gender and age and PC explaining more than 1% of the variance where stated. For the bron-chial biopsies, IL1RL1 expression was compared to the selected SNPs in 70 healthy subjects, with bronchi-al brushing IL1RL1 expression data being available in 72 out of the 77 hebronchi-althy persons.
ENCODE
We used data collected by the ENCODE Consortium60,61 to identify potential functional significance of the
associated SNPs within HaploReg v4.1.48,62 IL1RL1 TIR domain recombinant experiments
To examine differences in NF-Kβ signalling between TIR domain haplotypes post IL33 stimulation in the presence and absence of anti-IL33 or anti-ST2 we used Kruskal-Wallis test followed by Bonferroni post-hoc test. In the TLR experiments we tested two potential functional effects of the IL1RL1-b exon 11 hap-lotypes based on literature. First, we tested whether the two haphap-lotypes IL1RL1-b showed a differential suppressive effect on TLR2 and TLR4 signalling, as previously reported for IL1RL1-b. IL1RL1-b exon 11 risk and protective haplotypes were overexpressed in HeKBlue cells sensitive to either TLR2 stimulation with Pam3Cys or TLR4 stimulation with LPS.
Ethical Approval
The DAG and NORM cohort were approved by the Medical Ethics Committee of the University Medical Center Groningen. The Lung Tissue database study was approved by the ethics committees of the Institut universitaire de cardiologie et de pneumologie de Québec and the UBC-Providence Health Care Research Institute Ethics Board for Laval and UBC, respectively. The study protocol was consistent with the Research Code of the University Medical Center Groningen and Dutch national ethical and professional guidelines. In the AHBED dataset brushes were collected under ethics REC 08/H0406/189 (University of Leicester) and REC 08/H0407/1 (University of Nottingham). All participants in this study provided informed con-sent. GASP is a multicenter study under ethics GM129901, however also includes samples collected under local ethics from Nottingham (recruited 1990-2015), Belfast (recruited 2008-2009), Birmingham (2005-2014), Manchester (recruited 2008-(2005-2014), Southampton (recruited 2003-(2005-2014), Glasgow (recruited 2002-2014) and Leicester (recruited 2004-2015). All studies had appropriate local ethics approval.
Acknowledgments
This study was funded by an Asthma UK Grant to I.S., I.P.H., D.E.S., C.E.B. (AUK-PG-2013-188) and addition-al Asthma UK funding to I.S. and D.E.S (Grants 10/006 and 11/031). Genotyping in GASP was additionaddition-ally supported by Rosetrees Tust (Grant to I.S.), AirPROM and an MRC Strategic Award to I.P.H., M.D.T., L.V.W. and Professor David Strachan (MC_PC_12010). L.V.W holds a GSK/ British Lung Foundation Chair in Respi-ratory Research. Asthma UK also funded the GASP initiative (AUK-PG-2013-188). This work was part fund-ed by the NIHR Leicester Respiratory Biomfund-edical Centre. A.S. is supportfund-ed by the Manchester Biomfund-edical Research Centre. This study was also supported by a Lung Foundation of the Netherlands Dutch Lung Foundation (grant no. AF 95.05, AF 98.48 and no.AF3.2.09.081JU), the University Medical Center Gronin-gen and the Ubbo Emmius Fund.
